3rd issue of the Gradio newsletter! 🆕

Every two weeks, we’ll be sharing the latest news from the machine learning world, including the state-of-the-art machine learning models and demos that you can try right in your browser. Read on below 👇

• Gradio Version 2.7.5 ✅
• New Gradio Huggingface Spaces Demos ✅
  • First Order Motion Model 📺
  • GLIDE: Towards Photorealistic Image Generation and Editing with Text-Guided Diffusion Models 🖌️
  • MT3: Multi-Task Multitrack Music Transcription 🎵
  • JoJoGAN: One Shot Face Stylization 😀
  • PaddleSeg - Matting 🌇

First of all, if you haven’t already done so, consider taking 5 seconds and giving us a GitHub star to help our free, open-source library get more visible!

Gradio Version 2.7.5 ✅

We released Gradio version 2.7.5 with a lot of new features. Read the full release notes here, or we’ve summarized the key enhancements below: 🚀

Gradio version 2.7.5 includes several enhancements to the library: 🆕

1. Backend Migration to Starlette and FastAPI 🛠️
2. (BETA) Custom Flagging Callbacks and HuggingFaceDatasetSaver 🏁
3. Small Fixes 🔧

🆕 Huggingface Spaces 🚀

🤗 Spaces is out of beta, create your demos on Spaces using Gradio 🚀 Here are some good ones to check out.

First Order Motion Model

animate face images with a driving video

• Paper: https://arxiv.org/abs/2003.00196
• Spaces Demo: https://huggingface.co/spaces/abhishek/first-order-motion-model
• Spaces Code: https://huggingface.co/spaces/abhishek/first-order-motion-model/blob/main/app.py

glide-text2im

GLIDE: Towards Photorealistic Image Generation and Editing with Text-Guided Diffusion Models.

• paper: https://arxiv.org/abs/2112.10741
• Spaces Demo: https://huggingface.co/spaces/valhalla/glide-text2im
• Spaces Code: https://huggingface.co/spaces/valhalla/glide-text2im/blob/main/app.py

MT3: Multi-Task Multitrack Music Transcription

MT3 is a multi-instrument automatic music transcription model that uses the T5X framework.

• paper: https://arxiv.org/abs/2111.03017
• Spaces Demo: https://huggingface.co/spaces/akhaliq/MT3
• Spaces Code: https://huggingface.co/spaces/akhaliq/MT3/blob/main/app.py

JoJoGAN: One Shot Face Stylization

Only one single reference image is needed for training which takes about 1 minute on GPU in colab. After training, style can be applied to any input image

• paper: https://arxiv.org/abs/2112.11641
• Spaces Demo: https://huggingface.co/spaces/akhaliq/JoJoGAN
• Spaces Code: https://huggingface.co/spaces/akhaliq/JoJoGAN/blob/main/app.py

PaddleSeg - Matting

Easy-to-use image segmentation library with awesome pre-trained model zoo, supporting wide-range of practical tasks from research to industrial applications.

• Github: https://github.com/PaddlePaddle/PaddleSeg
• Spaces Demo: https://huggingface.co/spaces/DeepDrivePL/PaddleSeg-Matting
• Spaces Code: https://huggingface.co/spaces/DeepDrivePL/PaddleSeg-Matting/blob/main/app.py

Thanks for reading!

follow us on twitter for updates on gradio